Development Of Support Vector Machine Research Articles

Koroner Arter Hastalığı Sınıflandırılmasında Destek Vektör Makinelerinin Gri Kurt Optimizasyonuna Dayalı Özellik Seçim Yöntemi ile Geliştirilmesi

Makine öğrenmesi yöntemleri, büyük veri kümelerinin analiz edilmesine olanak sağlayarak koroner arter rahatsızlığı ve/veya buna benzer hastalık ve durumların tespit edilmesinde kullanılan etkili bir araçtır. Büyük veri kümelerinde işlem hızını ve sınıflandırma başarımını etkileyen gereksiz veya kararı olumsuz yönde etkileyen veriler bulunabilmektedir. Özellik seçim tekniklerinin uygulanması gereksiz verilerin ortadan kaldırılmasına olanak sağlamaktadır. Bu çalışmada, koroner arter hastalığını teşhis etmek amacıyla en uygun özellik alt kümesini belirlemek üzere yeni bir sınıflandırma yöntemi önerilmiştir. Önerilen yöntem, öznitelik seçimi ve sınıflandırma olmak üzere iki ana aşamadan oluşmaktadır. Önerilen yöntemin performans doğrulaması için Cleveland kalp hastalığı veri seti kullanılmıştır. İlk aşamada, en iyi özellikleri bulmak için gri kurt optimizasyonu (GWO) kullanılmıştır. Kullanılan veri setinde bulunan 13 parametre arasında 7 en etkili parametre seçilmiş ve sınıflandırma işlemi bu 7 parametre üzerinden gerçekleştirilmiştir. İkinci aşamada, GWO'nun uygunluk fonksiyonu, destek vektör makinesi (SVM) sınıflandırıcısı kullanılarak değerlendirilmiştir. Çalışmada belirlenen uygunluk fonksiyonları SVM’de kullanılan çekirdek matrislerin farklı varyasyonları ile değerlendirilmiştir. Bu aşamada en yüksek doğruluk elde edilen çekirdek matris belirlenmiştir. Deneysel sonuçlar, önerilen GWO-SVM'nin lineer çekirdek matris kullanılarak %95.91 doğrulukta, %95.64 duyarlılıkta ve %91.66 başarı ile mevcut çalışmalara kıyasla daha yüksek başarım sağlandığını göstermiştir

A joint learning framework for optimal feature extraction and multi-class SVM

In high-dimensional data classification, effectively extracting discriminative features while eliminating redundancy is crucial for enhancing the performances of classifiers, such as Support Vector Machine (SVM). However, previous studies have decoupled the process of feature extraction from the development of SVM, leading to suboptimal classification accuracy. To address this problem, we propose a novel joint learning framework that combines optimal feature extraction and multi-class SVM, incorporating a generalized regression form to learn a discriminative latent subspace. The projected data in this subspace are more likely to have a larger margin between different classes and align with the properties of the SVM classification mechanism, enhancing the overall classification performance. Three iterative algorithms were presented to obtain optimal solutions with guaranteed convergence, and theoretical analyses were also conducted to reveal their fundamental nature. The optimal linear projection subspace is equivalent to that obtained from Linear Discriminant Analysis (LDA) in some special cases. We conducted extensive experiments using diverse datasets to evaluate the performances of the proposed algorithms. Our algorithms achieved an accuracy improvement of up to 7.55% compared to other conventional methods.

Deciphering complex antibiotic resistance patterns in Helicobacter pylori through whole genome sequencing and machine learning.

Helicobacter pylori (H.pylori, Hp) affects billions of people worldwide. However, the emerging resistance of Hp to antibiotics challenges the effectiveness of current treatments. Investigating the genotype-phenotype connection for Hp using next-generation sequencing could enhance our understanding of this resistance. In this study, we analyzed 52 Hp strains collected from various hospitals. The susceptibility of these strains to five antibiotics was assessed using the agar dilution assay. Whole-genome sequencing was then performed to screen the antimicrobial resistance (AMR) genotypes of these Hp strains. To model the relationship between drug resistance and genotype, we employed univariate statistical tests, unsupervised machine learning, and supervised machine learning techniques, including the development of support vector machine models. Our models for predicting Amoxicillin resistance demonstrated 66% sensitivity and 100% specificity, while those for Clarithromycin resistance showed 100% sensitivity and 100% specificity. These results outperformed the known resistance sites for Amoxicillin (A1834G) and Clarithromycin (A2147), which had sensitivities of 22.2% and 87%, and specificities of 100% and 96%, respectively. Our study demonstrates that predictive modeling using supervised learning algorithms with feature selection can yield diagnostic models with higher predictive power compared to models relying on single single-nucleotide polymorphism (SNP) sites. This approach significantly contributes to enhancing the precision and effectiveness of antibiotic treatment strategies for Hp infections. The application of whole-genome sequencing for Hp presents a promising pathway for advancing personalized medicine in this context.

Analysis and Prediction of Students’ Psychological Characteristics of Law Breaking and Crime Based on Support Vector Machine Algorithm

Abstract In recent years, some people with high intelligence and high quality have violated the law and even committed crimes, which has ruined their bright future. It is really sad. A school is a place where teenagers learn skills and improve their quality. It should be a pure land. The frequent occurrence of juvenile students’ crimes seriously affects the normal teaching order and endangers the safety of students’ lives and property. With the rapid development of support vector machines (Abbreviation: SVM), it has also become one of the main data fusion technologies. SVM is a creative machine learning method based on statistical learning theory. It is a learning method specialized in finite sample prediction. This paper studies the psychological characteristics of students’ crimes based on the SVM algorithm. From the aspect of proportion, the result of the SVM algorithm is ideal, and the proportion of the training set is up to 33.35%. The above algorithm has good expansibility for such sample data, but when the parameters are not appropriate, the proportion is only more than 20.12%, which once again explains the importance of the SVM algorithm. Only through the SVM algorithm to improve the efforts of the whole society, can we put an end to the generation of criminal psychology of higher vocational students from the source and cultivate more high-quality talents for the country.

Research of melting state identification and process performance based on selective laser melting acoustic signals

An acoustic signal acquisition experiment platform was constructed to gather the acoustic signals throughout the formation of 35 single-tracks of a 120 mm length copper-tin alloy in order to monitor and precisely manage the selective laser melting (SLM) forming process and enhance overall quality. The monitoring of the SLM forming process includes the analysis of the time and frequency domains, the extraction of the SLM process features using linear prediction techniques, and the development of support vector machine (SVM) model, back-propagation (BP) neural network models, and convolutional neural network models. The results show that the over-melted state can be identified by extracting time and frequency-domain features over a given range, but the normal and unmelted states are difficult to distinguish. The convolutional neural network model had a recognition rate of 99%, the BP neural network had an effective recognition rate of 90%, and the SVM model had a combined classification rate of 83.14% for the three states after optimization. In contrast, the convolutional neural network model performs best in monitoring and offers a framework and point of reference for acoustic signal analysis and online SLM quality monitoring.

Assessment of circular-bored twin tunnel (CBTT) performance using soft computing methods

Circular-bored tunnel (CBTT) machines are used as an alternative to drill and for blasting in rock and conventional "hand mining" in soil. The penetration rate (PR) prediction is a difficult issue plus the related analysis between CBTT and rock mass. Circular tunnels are high proposed for general projects. Experimental and analytical models in analyzing CBTT penetration ratio could not provide a reliable and valid prediction to civil projects and mining because of the nonlinear problems, related risks and high costs. This research provides a novel study of using three algorithms as support vector machine (SVM), grey wolf optimizer (GWO) and k-nearest neighbor (KNN) to predict the PR of CBTT. For this purpose, doing empirical analysis could be time consuming, lead to high cost, inaccuracy of results, and high error percentages. Accordingly, this study, by doing numerical analysis, has tried to accurately predict the PR of CBTT in terms of uniaxial compressive strength and rock quality designation were used as input data. It was followed by the development of SVM, KNN, and GWO to obtain the best predictive model. Accordingly, SVM with the best RMSE value of (0.183), R2 value (0.914) and MAE value of (0.139) was selected as the best model among other two ones in this research, meaning that SVM could predict the PR of CBTT with more accuracy and reliably. Thus, SVM model identified uniaxial compressive strength (0.2) as a significant parameter in CBTT prediction and penetration ratio.

Early Detection of Ganoderma boninense in Oil Palm Seedlings Using Support Vector Machines

Ganodermaboninense (G. boninense) is a fungus that causes one of the most destructive diseases in oil palm plantations in Southeast Asia called basal stem rot (BSR), resulting in annual losses of up to USD 500 million. The G. boninense infects both mature trees and seedlings. The current practice of detection still depends on manual inspection by a human expert every two weeks. This study aimed to detect early G. boninense infections using visible-near infrared (VIS-NIR) hyperspectral images where there are no BSR symptoms present. Twenty-eight samples of oil palm seedlings at five months old were used whereby 15 of them were inoculated with the G. boninense pathogen. Five months later, spectral reflectance oil palm leaflets taken from fronds 1 (F1) and 2 (F2) were obtained from the VIS-NIR hyperspectral images. The significant bands were identified based on the high separation between uninoculated (U) and inoculated (I) seedlings. The results indicate that the differences were evidently seen in the NIR spectrum. The bands were later used as input parameters for the development of Support Vector Machine (SVM) classification models, and these bands were optimized according to the classification accuracy achieved by the classifiers. It was observed that the U and I seedlings were excellently classified with 100% accuracy using 35 bands and 18 bands of F1. However, the combination of F1 and F2 (F12) gave better accuracy than F2 and almost similar to F1 for specific classifiers. This finding will provide an advantage when using aerial images where there is no need to separate F1 and F2 during the data pre-processing stage.

Robust statistics-based support vector machine and its variants: a survey

Support vector machines (SVMs) are versatile learning models which are used for both classification and regression. Several authors have reported successful applications of SVM in a wide range of fields. With the continuous growth and development in machine learning using SVM, it was observed that SVM also has some limitations. This paper focuses on limitation regarding its boundary, i.e., sensitivity to noise or outliers in the dataset. Researchers have proposed many variants and extensions of SVM to make it robust. This paper gives an overview of the developments in the field of robust statistics in support vector machines and its variants. This paper includes an up to date survey of the research development in the field of robustness in SVM and its extensions. It also includes a discussion part which not only discusses the pros and cons of the proposed approaches but also highlights some important future directions in it. This paper would be helpful for researchers working in the field of robust statistics as well as supervised machine learning. This study would also encourage the researchers to work further in the development of SVM and even its variants to improve them.

Prediction of Scour Depth below River Pipeline using Support Vector Machine

In this paper, the depth of scouring phenomenon below the pipelines across rivers was predicted using Support Vector Machine (SVM). To this end, the related dataset was collected from literature. Performance of SVM was evaluated via calculation of error indices such as coefficient of determination (R2) and Root Mean Square of Error (RMSE). The accuracy of SVM was compared with artificial neural network (ANN) and Adaptive Neuro fuzzy Inference Systems (ANFIS). To find out the most effective parameters on scouring depth, a sensitivity analysis was conducted on ANN, ANFIS, and SVM. During the development of SVM, it was found that this model with R2= 0.94 and RMSE = 0.103 in testing stage has a suitable performance for predicting the scouring depth below the river pipeline. Assessing kernel functions showed that radial basis function has the best outcomes. Comparing the accuracy of SVM with ANN and SVM showed that the accuracy of SVM is a bit better than ANN with R2 = 0.89 and RMSE = 0.12 and ANFIS with R2 = 0.92 and RMSE = 0.13. Sensitivity analysis showed that e/D, τ* and Fr are the most effective parameters for predicting the scouring depth below the pipeline.

The Early Psychosis Screener for Internet (EPSI)-SR: Predicting 12 month psychotic conversion using machine learning

IntroductionA faster and more accurate self-report screener for early psychosis is needed to promote early identification and intervention. MethodsSelf-report Likert-scale survey items were administered to individuals being screened with the Structured Interview for Psychosis-risk Syndromes (SIPS) and followed at eight early psychosis clinics. An a priori analytic plan included Spectral Clustering Analysis to reduce the item pool, followed by development of Support Vector Machine (SVM) classifiers. ResultsThe cross-validated positive predictive value (PPV) of the EPSI at the default cut-off (76.5%) exceeded that of the clinician-administered SIPS (68.5%) at separating individuals who would not convert to psychosis within 12 months from those who either would convert within 12 months or who had already experienced a first episode psychosis (FEP). When used in tandem with the SIPS on clinical high risk participants, the EPSI increased the combined PPV to 86.6%. The SVM classified as FEP/converters only 1% of individuals in non-clinical and 4% of clinical low risk populations. Sensitivity of the EPSI, however, was 51% at the default cut-off. DiscussionThe EPSI identifies, comparably to the SIPS but in less time and with fewer resources, individuals who are either at very high risk to develop a psychotic disorder within 12 months or who are already psychotic. At its default cut-off, EPSI misses 49% of current or future psychotic cases. The cut-off can, however, be adjusted based on purpose. The EPSI is the first validated assessment to predict 12-month psychotic conversion. An online screening system, www.eps.telesage.org, is under development.

Daily River Flow Forecasting with Hybrid Support Vector Machine – Particle Swarm Optimization

The application of artificial intelligence techniques for river flow forecasting can further improve the management of water resources and flood prevention. This study concerns the development of support vector machine (SVM) based model and its hybridization with particle swarm optimization (PSO) to forecast short term daily river flow at Upper Bertam Catchment located in Cameron Highland, Malaysia. Ten years duration of historical rainfall, antecedent river flow data and various meteorology parameters data from 2003 to 2012 are used in this study. Four SVM based models are proposed which are SVM1, SVM2, SVM-PSO1 and SVM-PSO2 to forecast 1 to 7 day ahead of river flow. SVM1 and SVM-PSO1 are the models with historical rainfall and antecedent river flow as its input, while SVM2 and SVM-PSO2 are the models with historical rainfall, antecedent river flow data and additional meteorological parameters as input. The performances of the proposed model are measured in term of RMSE and R2. It is found that, SVM2 outperformed SVM1 and SVM-PSO2 outperformed SVM-PSO1 which meant the additional meteorology parameters used as input to the proposed models significantly affect the model performances. Hybrid models SVM-PSO1 and SVM-PSO2 yield higher performances as compared to SVM1 and SVM2. It is found that hybrid models are more effective in forecasting river flow at 1 to 7 day ahead at the study area.

In-process tool condition monitoring in compliant abrasive belt grinding process using support vector machine and genetic algorithm

Industrial interest in tool condition monitoring for compliant coated abrasives has significantly augmented in recent years as unlike other abrasive machining processes the grains are not regenerated. Tool life is a significant criterion in coated abrasive machining since deterioration of abrasive grains increases the surface irregularity and adversely affects the finishing quality. Predicting tool life in real time for coated abrasives not only helps to optimise the utilisation of the tool’s life cycle but also secures the surface quality of finished components. This paper describes the evolution of the abrasive grain degradation in the belt tool with process time and also the development of Support Vector Machine (SVM) and Genetic Algorithm (GA) based predictive classification model for in-process sensing of abrasive belt wear for robotized abrasive belt grinding process. With this tool condition monitoring predicting system, the effectiveness of the belt and the surface integrity of the material is secure. The analysis of sensor signals generated by the accelerometer, Acoustic Emission (AE) sensor and force sensor during machining is proposed as a technique for detecting belt tool life states. Various time and frequency domain features are extracted from sensor signals obtained from the accelerometer, acoustic sensor and force sensor mounted on the belt grinding setup. The time and frequency domain features extracted from the signals are simultaneously optimised to obtain a subset with fewer input features using a GA. The classification accuracy of the k-Nearest Neighbour (kNN) technique is used as the fitness function for the GA. The subset features extracted from the signals are used to train the SVM in MATLAB. An experimental investigation using four different conditions of tool states is introduced to the SVM and GA for the prediction and classification. By the experimental results, this research proves that the proposed SVM based in-process tool condition monitoring model has a high accuracy rate for predicting abrasive belt condition states.

Prediction of discharge coefficient of cylindrical weir–gate using GMDH-PSO

Cylindrical weir-gate consists of a weir and a gate has been proposed for flow measurement in water engineering projects. Discharge coefficient of weir-gate is a fundamental parameter to evaluate its hydraulic efficiency. In this research, the group method of data handling (GMDH) trained by particle swarm optimization technique was used to predict the discharge coefficient of cylindrical weir-gate. The performances of prepared GMDH model were compared with multi-layer perceptron neural network (MLPNN) and support vector machine (SVM) that developed to this end, as well. Results indicated that all developed models have suitable performance, however; the SVM model was a bit more accurate. Observing the obtained structure of GMDH model showed that the upstream Froude number and ratio of opening height of gate to the diameter of cylindrical weir-gate are the most effective parameters on discharge coefficient. During the development of SVM and MLPNN models, it was found that the radial basis function as kernel function and hyperbolic tangent sigmoid as transfer function have a better accuracy compared to other tested functions.

PAC-Bayes bounds for twin support vector machines

Twin support vector machines are regarded as a milestone in the development of support vector machines. Compared to standard support vector machines, they learn two nonparallel hyperplanes rather than one as in standard support vector machines for binary classification, and work faster and sometimes perform better than support vector machines. One of the reasons that support vector machines are widely used is that they are supported by strong statistical learning theory. However, relatively little is known about the theoretical analysis of twin support vector machines. As recent tightest bounds for practical applications, PAC-Bayes bound and prior PAC-Bayes bound are based on a prior and posterior over the distribution of classifiers. In this paper, we study twin support vector machines from a theoretical perspective and use the PAC-Bayes bound and prior PAC-Bayes bound to measure the generalization error bound of twin support vector machines. Experimental results on real-world datasets show better predictive capabilities of the PAC-Bayes bound and prior PAC-Bayes bound for twin support vector machines compared to the PAC-Bayes bound and the prior PAC-Bayes bound for support vector machines.

Actualité en neuro-imagerie du Trouble Déficit d’Attention/Hyperactivité

An increasing number of neuroimaging (in particular, magnetic resonance imaging [MRI]) studies has been published in the last two decades with the aim to elucidate the brain correlates of Attention-Deficit/Hyperactivity Disorder (ADHD). The amount of available evidence has allowed researchers to pool such body of studies in meta-analyses, which provide more reliable information than single and often underpowered studies. Meta-analytic evidence from structural MRI studies shows that, compared to non-ADHD controls, individuals with ADHD present with significant differences in the volume of fronto-parietal, striatal, thalamic and cerebellar regions, involved in inhibition, self-adjustment and goal-directed behaviors, functions typically impaired in ADHD. Converging evidence from diffusion tensor imaging studies also shows ADHD-related white matter alterations in fronto-striatal-cerebellar circuits as well as in parieto-temporal-occipital regions. Evidence from structural studies has been complemented by task-based functional MRI studies pointing to dysfunctions in the same brain regions. The relatively recent introduction of resting-state MRI has allowed the detection of complex patterns of dysfunctional interactions, at rest, among several brain networks, including the default and task-positive networks, in individuals with ADHD. To date, most of the neuroimaging literature in ADHD has focused on group comparisons; therefore, its results are not applicable at the single-patient level. However, the recent development of support vector machine and similar analytical approaches promises to turn the field towards useful application in terms of diagnosis and prognosis at the single-patient level, thus being informative for daily clinical decisions.

Support Vector Machine on Fluorescence Landscapes for Breast Cancer Diagnostics

Excitation-emission matrices (EEM) and total synchronous fluorescence spectra (SFS) of normal and malignant breast tissue specimens are measured in UV-VIS spectral region to serve as data inputs in development of Support Vector Machine (SVM) based breast cancer diagnostics tool. Various input data combinations are tested for classification accuracy using SVM prediction against histopathology findings to discover the best combination regarding diagnostics sensitivity and specificity. It is shown that with EEM data SVM provided 67% sensitivity and 62% specificity diagnostics. With SFS data SVM provided 100% sensitivity and specificity for a several input data combinations. Among these combinations those that require minimal data inputs are identified.

Online Learning Algorithms of Direct Support Vector Machine

With the development of support vector machine, online learning became important in many areas.-Direct support vector machine (DSVM) was a new deformation algorithm for standard support vector-machine(SVM) which had fewer computation complexity for training and prediction. Based on the-mathematical mode of DSVM for classification and regression, as well as using the property of block-matrix and kernel function matrix, we deduced online learning algorithms of DSVM according to time-window theoretically. By using Mackey-Glass chaotic time series for simulation, the results show that-the algorithms are workable and predictive values are in good accordance with original values. Our-work favours SVM's applications where online learning is necessary.

Polarimetric radar target classification using support vector machines

This paper reports on the development of support vector machines for polarimetric radar target classification. Automatic classification of radar signatures of man-made objects has been shown to be improved noticeably by optimum selection of transmitting and receiving states of polarization. However, the optimum polarization states may be target-dependent. Thus the need for an optimum classifier to correctly label radar signatures associated with different targets when sensed at differing pairs of transmitting and receiving polarization states. To generate radar signatures of targets at various transmit-receive pairs of polarization angles from vertical and horizontal states, the technique of polarization synthesis was applied. Then statistical attributes from each radar signature were used for its representation. Finally, support vector machines, using a number of kernels such as linear functions, polynomials, Gaussian radial basis functions, exponential basis functions, linear splines, and basis spline functions, were developed and used on real fully polarimetric radar data. The results indicate that a small subset of polarization angles are sufficient for generating signatures needed for training a classifier for optimal separation of polarimetric-diverse signatures. Moreover, the classification performance in terms of receiver operating characteristic curves shows that Gaussian kernels outperform other kernels that we have used.

Development of SVM based ECG Pattern Recognition Technique

This paper presents simple and reliable method for the detection of QRS complexes (cardiac beat) in single lead Electrocardiograms (ECG) using Support Vector Machine (SVM). Two different preprocessing techniques are applied for the generation of features. First involves digital filtering to remove base line wander and power line interference while the second involves entropy criterion for feature generation. SVM is used as a classifier to classify QRS and non-QRS regions. The results of the validation of the method on the standard CSE ECG database are also presented. Detection rate of 99.79 % is achieved. The percentage of false positive and false negative is 0.86% and 0.21% respectively. The proposed method functions reliably even under the condition of poor signal quality of the ECG. The detection rate depends strongly on the quality of training, data representation and the mathematical basis of the classifier.

Kernel Nearest-Neighbor Algorithm

The ‘kernel approach’ has attracted great attention with the development of support vector machine (SVM) and has been studied in a general way. It offers an alternative solution to increase the computational power of linear learning machines by mapping data into a high dimensional feature space. This ‘approach’ is extended to the well-known nearest-neighbor algorithm in this paper. It can be realized by substitution of a kernel distance metric for the original one in Hilbert space, and the corresponding algorithm is called kernel nearest-neighbor algorithm. Three data sets, an artificial data set, BUPA liver disorders database and USPS database, were used for testing. Kernel nearest-neighbor algorithm was compared with conventional nearest-neighbor algorithm and SVM Experiments show that kernel nearest-neighbor algorithm is more powerful than conventional nearest-neighbor algorithm, and it can compete with SVM.